120
L. Benyahya et al.
|
Frequency analysis of seasonal extreme precipitation in southern Quebec (Canada)
Hydrology Research
|
45.1
|
2014
functions in water resources engineering (Montana type; i.e.,
statistics is greater than that of the gridded/simulated data
Mohymont & Demarée ).
products, which means that these products systematically underestimate the observed variance. This result can be
i(d, T ) ¼
a(T) dη(T )
(2)
where a and η are coefficients to be estimated.
the consequence of the so-called ‘smoothing effect’ often associated with interpolation (or the search neighborhood). Moreover, with respect to the mean, it is clear that, for both seasons, CS shows values close to the observations, while ANUSPLIN and CRCM tend to show negative (underestimation) bias. For the minimum SMP, in winter, the two
RESULTS AND DISCUSSION
gridded products (CS and ANUSPLIN) overestimate the observed values, while the CRCM underestimates them. In
The SMP data over the 83 grid points obtained from each
summer months, the gridded/simulated SMP values repro-
dataset (CS, ANUSPLIN, and CRCM4.1.1) are first com-
duced the observed data quite well; while there is a slight
pared with the observed meteorological stations (69) over
overestimation from the CS values. For the maximum
the whole study area and the entire time window (1961–
SMP, in winter, the values are also quite well reproduced
1999). The summary of various statistics (i.e., minimum,
by all gridded/simulated data, except for the ANUSPLIN
maximum, mean, skewness, and coefficient of variation)
which underestimates the observed values. In summer, this
for 1-day SMP is presented in Figure 2. For both summer
location parameter is systematically underestimated by all
and winter seasons, the observed variability of all the
gridded/simulated values, especially with the CRCM
Figure 2
|
Box plots of statistics of seasonal 1-day maximum precipitation of observed meteorological stations, CS, ANUSPLIN, and CRCM 4.1.1 (1961–1999) for winter season (upper panels) and summer season (lower panels).